Heterogeneous ScaLAPACK (HeteroScaLAPACK) 1.0.2-BETA

This release contains performance improvements.
The auxiliary routines (_info functions), especially for heterogeneous ScaLAPACK routines, were incurring large execution times due to expensive Cblacs_gridinit and Cblacs_gridmap calls. They have been optimized.

The evaluation of process arrangements in HMPI_Group_heuristic_pauto_d_create is now done by processors in parallel rather than processes. This is because in some cases all the process arrangements may be allocated to the processes on a single processor leading to performance hit.